Follistatin and Noggin Are Excluded from the Zebrafish Organizer

Home , Decapentaplegic, Follistatin, NODAL, Noggin (protein)

DEVELOPMENTAL BIOLOGY 204, 488–507 (1998) ARTICLE NO. DB989003 Follistatin and Noggin Are Excluded from the Zebraﬁsh Organizer

Hermann Bauer,* Andrea Meier,* Marc Hild,* Scott Stachel,†,1 Aris Economides,‡ Dennis Hazelett,† Richard M. Harland,† and Matthias Hammerschmidt*,2 *Max-Planck Institut fu¨r Immunbiologie, Stu¨beweg 51, 79108 Freiburg, Germany; †Department of Molecular and Cell Biology, University of California, 401 Barker Hall 3204, Berkeley, California 94720-3204; and ‡Regeneron Pharmaceuticals, Inc., 777 Old Saw Mill River Road, Tarrytown, New York 10591-6707

The patterning activity of the Spemann organizer in early amphibian embryos has been characterized by a number of organizer-specific secreted proteins including Chordin, Noggin, and Follistatin, which all share the same inductive properties. They can neuralize ectoderm and dorsalize ventral mesoderm by blocking the ventralizing signals Bmp2 and Bmp4. In the zebrafish, null mutations in the chordin gene, named chordino, lead to a severe reduction of organizer activity, indicating that Chordino is an essential, but not the only, inductive signal generated by the zebrafish organizer. A second gene required for zebrafish organizer function is mercedes, but the molecular nature of its product is not known as yet. To investigate whether and how Follistatin and Noggin are involved in dorsoventral (D-V) patterning of the zebrafish embryo, we have now isolated and characterized their zebrafish homologues. Overexpression studies demonstrate that both proteins have the same dorsalizing properties as their Xenopus homologues. However, unlike the Xenopus genes, zebrafish follistatin and noggin are not expressed in the organizer region, nor are they linked to the mercedes mutation. Expression of both genes starts at midgastrula stages. While no patterned noggin expression was detectable by in situ hybridization during gastrulation stages, later expression is confined to presumptive cartilage cells in the branchial arches and the neurocranium and to proximal regions of the pectoral fin buds. follistatin transcripts in gastrulating embryos are confined to anterior paraxial regions, which give rise to head mesoderm and the first five somites. The dorsolateral extent of this expression domain is regulated by Bmp2b, Chordino, and Follistatin itself. In addition, transient expression was observed in a subset of cells in the posterior notochord anlage. Later, follistatin is expressed in brain, eyes, and somites. Comparison of the spatiotemporal expression pattern of follistatin and noggin with those of bmp2b and bmp4 and overexpression studies suggest that Noggin and Follistatin may function as Bmp antagonists in later processes of zebrafish development, including late phases of D-V patterning, to refine the early pattern set up by the interaction of Chordino and Bmp2/4. It thus appears that many, but not all, aspects of early dorsoventral patterning are shared among different vertebrate species. © 1998 Academic Press

INTRODUCTION step process involving inductive processes driven by different maternally and zygotically supplied signals (for review, Studies in Amphibia have revealed that axis formation see Harland and Gerhart, 1997). Maternal signals induce and early dorsoventral (D-V) patterning represent a multi- dorsal and ventral mesoderm in the equatorial zone of the early blastula embryo. This coarse early pattern is then Sequence data for this article have been deposited with the refined by zygotic signals. Proteins secreted by the dorsal GenBank Data Library under Accession Nos. AF 084948 (Follista- mesoderm, also called the Spemann organizer (Spemann tin) and AF 084949 (Noggin). and Mangold, 1924), induce neural specification in dorsal 1 Present address: Center for the Application of Molecular Biol- animal regions which would otherwise give rise to epider- ogy to International Agriculture, GPO Box 3200, Canberra ACT 2601, Australia. mal derivatives. In addition, the same signals lead to a 2 To whom correspondence should be addressed. Telephone: ϩ49- dorsalization of initially ventrally specified mesodermal 761-5108 495. Fax: ϩ49-761-5108 358. E-mail: hammerschmid@ cells in lateral regions (see for review, Lemaire and Kodjaba- immunbio.mpg.de. chian, 1996; Harland and Gerhart, 1997). Five proteins

0012-1606/98 $25.00 Copyright © 1998 by Academic Press 488 All rights of reproduction in any form reserved. Follistatin and Noggin in Zebrafish 489 which are expressed in the organizer region of Xenopus described Bmp2/4 antagonists, Follistatin and Noggin? In embryos and which display the inductive properties as- contrast to chordin, no Drosophila follistatin and noggin signed to the Spemann organizer have been identified: homologues have been reported, although Xenopus noggin Chordin (Sasai et al., 1994), Noggin (Smith and Harland, expression was shown to affect D-V patterning by antago- 1992), Follistatin (Hemmati-Brivanlou et al., 1994), Xnr3 nizing Dpp when expressed in Drosophila embryos ( Holley (Smith et al., 1995), and Cerberus (Bouwmeester et al., et al., 1996). In Xenopus, zygotic noggin expression starts at 1996). Among these, Follistatin was originally described as late blastula stages in the dorsal marginal zone and persists an inhibitor of Activin (Nakamura et al., 1989; Hemmati- throughout gastrulation in the prechordal plate and the Brivanlou et al., 1994). In misexpression experiments in presumptive notochord, both derivatives of the Spemann Xenopus embryos, all five proteins can dorsalize ventral organizer. At later stages, noggin expression is initiated at mesoderm (Smith and Harland, 1992; Smith et al., 1993, several new sites, including the roof plate of the neural tube 1995; Sasai et al., 1994; Fainsod et al., 1997; Hsu et al., and skeletogenic cells in the branchial arches (Smith and 1998) and induce neural specification in the absence of Harland, 1992). In the mouse, noggin is expressed in similar mesoderm (Lamb et al., 1993; Hemmati-Brivanlou et al., tissues, namely the node and its axial mesoderm deriva- 1994; Sasai et al., 1995; Hansen et al., 1997). Biochemical tives, as well as the roof plate and condensing cartilage studies indicate that they fulfill their dorsalizing and (McMahon et al., 1998). Mutation of noggin in the mouse neural-inducing activity indirectly via the inhibition of the appears to have no effect on node function, though subse- ventralizing Bone Morphogenetic Proteins (Bmps), mem- quent patterning of both the neural tube and the somites is bers of the TGF␤ growth factor family which appear to aberrant. function as instructive D-V patterning molecules, deter- Xenopus follistatin expression starts slightly later than mining in a dose-dependent fashion positional identities that of noggin. At the onset of gastrulation, follistatin RNA along the D-V axis of the early Xenopus embryo (Dosch et is detected in a few cells of the Spemann organizer. During al., 1997); Chordin, Noggin, Cerberus, and possibly Follista- gastrulation, follistatin expression continues in the pre- tin bind Bmp2 and Bmp4 on the dorsal side of the embryo, chordal plate and the anterior portion of the notochord thereby preventing binding of the Bmp proteins to their anlage. Beginning at early neurula stages, follistatin expres- receptor (Piccolo et al., 1996; Zimmerman and Harland, sion is initiated at new sites in the head mesoderm; the 1996; Fainsod et al., 1997; Hsu et al., 1998). hypochord; the pronephros; the eyes; the fore-, mid-, and Evidence revealing the requirement of Bmp2b and one of hindbrain; and the midbrain–hindbrain junction (Hemmati- its antagonists, Chordin, during early D-V patterning has Brivanlou et al., 1994). In the mouse, follistatin expression recently come from mutant studies in the zebrafish. Here, pattern is different; embryonic expression of follistatin first large-scale mutant screens (Driever et al., 1996; Haffter et occurs in the primitive streak, followed by expression in al., 1996) have led to the identification of six complemen- head mesoderm, somites, and specific rhombomeres of the tation groups defining six genes required for ventral devel- hindbrain, and later in midbrain and diencephalon (Albano opment (Mullins et al., 1996) and two complementation et al., 1994; Feijen et al., 1994). No expression was reported groups defining two genes required for organizer-dependent in the node, the mouse equivalent of the Spemann organizer dorsal development (Hammerschmidt et al., 1996a). Mo- (Beddington, 1994), or the notochord (Albano et al., 1994; lecular analyses showed that the phenotype of the strongest Feijen et al., 1994). Consistent with these results, follistatin of the dorsalized mutants, swirl, is caused by null muta- knockout mice display later defects, e.g., in muscle and tions in the zebrafish bmp2b gene (Kishimoto et al., 1997; skeleton, and no neural phenotype or defects in early D-V Nikaido et al., 1997; Martıńez-Barbera´ et al., 1997), while patterning are detected (Matzuk et al., 1995), calling into the ventralization observed in dino mutants is caused by a question whether follistatin is involved in the regulation of null mutation in the zebrafish chordin gene (Schulte- early D-V pattern formation of the mouse embryo. Merker et al., 1997; Fisher et al., 1997). In light of this For a comparative analysis of the roles of Follistatin and relationship, the dino mutant and the zebrafish chordin Noggin in early vertebrate development, we have cloned gene were renamed chordino. Injection studies and double- their zebrafish homologues; examined their spatiotemporal mutant analyses revealed that bmp2b/swirl is epistatic to expression pattern in comparison to that of the other chordino, consistent with the evidence that Chordin acts as known regulators of D-V patterning, bmp2b, bmp4 (Ni- an inhibitor of Bmp2 (Hammerschmidt et al., 1996b). Inter- kaido et al., 1997), and chordino (Schulte-Merker et al., estingly, in the fruitfly Drosophila melanogaster, structural 1997; Miller-Bertoglio et al., 1997); and characterized some homologues of Chordin and Bmp2/4, called Short gastrula- of their functional properties. We failed to detect any sign of tion and Decapentaplegic, respectively, have been shown to noggin or follistatin expression in zebrafish blastula or early regulate embryonic D-V patterning, suggesting that the gastrula embryos, when organizer signaling is occurring. At regulation of early D-V patterning may be conserved be- later stages, follistatin and noggin are expressed in distinct tween invertebrates and vertebrates (see for review, Jones domains that are often located in close proximity to expres- and Smith, 1995; Holley et al., 1995; De Robertis and Sasai, sion domains of bmp2b and bmp4. Our data suggest that 1996; Holley and Ferguson, 1997). Follistatin and Noggin may act as inhibitors of Bmps at Does this conservation also apply to the other two later stages of development, but are not involved in orga-

Copyright © 1998 by Academic Press. All rights of reproduction in any form reserved. 490 Bauer et al. nizer function or early phases of dorsoventral pattern for- run in Mops buffer. After capillary transfer onto Hybond-Nϩ mation. membrane in 20ϫ SSC, RNA was crosslinked to the filter by UV treatment in a Stratalinker (Stratagene), and filters were baked at 80°C for 2 h. Hybridization to detect follistatin transcripts was MATERIALS AND METHODS carried out overnight with the radioactively labeled entire insert of pBS-zfol 3Ј, supplemented with calf thymus DNA, at 42°C in 5ϫ SSC, 5ϫ Denhardt’s, 0.5% SDS, 100 mM sodium phosphate, pH Cloning of Zebrafish follistatin and noggin 6.8, 50% formamide. Blots were washed in 2ϫ SSC, 0.1% SDS, at For cloning of the zebrafish follistatin gene, 4 ϫ 105 phage clones room temperature and in 0.1ϫ SSC, 0.1% SDS at 65°C and were of a Lambda ZAP (Stratagene) gastrula-stage cDNA library (kind exposed on Kodak BIOMAX MS supersensitive X-ray film at Ϫ80°C gift from D. Grunwald) were screened with a radioactively labeled for 2–5 days. 859-bp BamHI–ApaI fragment of Xenopus follistatin which contains the major part of the conserved Follistatin module region. RNase Protection Assays Filters were hybridized overnight at 55°C in 6ϫ SSC, 5ϫ Den- To determine the temporal expression pattern of noggin mRNA, hardt’s solution, 1% SDS and washed at 60°C in 2ϫ SSC, 0.1% we performed RNase protection assays (Melton, 1984) using gel- SDS. Recombinant Bluescript plasmids were recovered from purified probes made from high-specific-activity [32P]UTP (6000 plaque-purified phage clones by in vivo excision using the Ex Assist Ci/mmol; NEN). Interference-Resistant Helper Phage kit (Stratagene), following the supplier’s instructions. Five different clones were obtained and sequenced using an ABImed automatic sequencer. One of the RT-PCR Analysis clones, named pBS-zfol3Ј, showed homology to published follista- Stage-specific RNA was isolated as described for Northern anal- tin sequences. It consists of 684 bp of coding region and 3Ј UTR but ysis. For RT-PCR, 10 ␮g of total RNA was subjected to digestion lacks the 5Ј end. The insert of this clone was used to screen 8 ϫ 105 with RNase-free DNase I (Boehringer, 50 units) in the presence of clones of a randomly-primed and an oligo(dT)-primed sphere-to-5- 100 units of RNasin (Promega) to degrade any genomic DNA. After somite-stage cDNA library (kind gift from J. Campos Ortega) under phenol–chloroform and chloroform extraction and precipitation stringent conditions, leading to the isolation of two independent with 1 vol of 2-propanol, RNA was redissolved and used for reverse clones with a full-length coding region (hybridization at 68°C in transcription. One microgram of RNA was incubated with 30 pmol 300 mM sodium phosphate, pH 6.8, 7% SDS, 10 mM EDTA; of random primers (Gibco) at 65°C for 10 min, chilled on ice, and washes at 70°C in 40 mM sodium phosphate, pH 6.8, 1% SDS; supplemented with 2 ␮l of 100 mM DTT, 2 ␮l of dNTPs (10 mM Church and Gilbert, 1984). The longest of the two clones, desig- each, Pharmacia), 20 units of RNasin, 4 ␮l of reverse transcription nated pBS-folT1, contains an insert of approximately 1.6 kb, with buffer (Promega), and 50 units of Expand reverse transcriptase 53 bp of 5Ј UTR, 966 bp of coding region, and approximately 600 bp (Boehringer) to a final reaction volume of 20 ␮l. After1hat42°C, of 3Ј UTR. 1 ␮l of these reactions was used as a template in 50-␮l PCRs with The noggin gene was cloned from a genomic library by standard 30 pmol of the respective PCR primers. PCR conditions were 95°C low-stringency hybridization using a Xenopus noggin cDNA. for 3 min–20ϫ (95°C for 30 s–60°C for 30 s–72°C for 45 s)–72°C for Genomic DNA from 72-h (postfertilization) zebrafish embryos of 2 min. For follistatin, the oligonucleotide CGATGATATCCACT- “wild-type” fish (Ekkwill Tropical Fish Breeders, Gibsonton, FL) GCAGTGCAGG was used as sense primer and the oligonucleotide was partially digested with Sau3A to an average size of 20 kb, the GTTTTGGGAAGGAGGTTGCTG as antisense primer to amplify first two nucleotides of the Sau3A ends were filled in, and the DNA a 274-bp fragment. noggin was monitored using the primers AC- was cloned into XhoI-digested and partially filled-in Lambda FIX II CGTGCTCATTTTCTCCCTCGGG (sense) and GACCAGAGC- (Stratagene). The ligation was packaged with Gigapack II XL CACAGCTGAAGCC (antisense), which amplify a 391-bp frag- Packaging Extract (Stratagene), which preferentially size selects for ment. To confirm that comparable amounts of template were used 47- to 51-kb recombinants (18- to 22-kb inserts). A primary library for the different stages, mRNA of the translation factor EF1␣ of 1.8 ϫ 106 clones was amplified in SRB/P2 to a titer of 4 ϫ 109 (Nordness et al., 1994) was quantified in parallel PCRs with the pfu/ml. Five independent clones contained overlapping segments primers TCACCCTGGGAGTGAAACAGC (sense) and ACTTG- of the same locus. noggin was not found in extensive screening of CAGGCGATGTGAGCAG (antisense), leading to a 692-bp frag- gastrula- or neurula stage-cDNA libraries (kind gift from D. Grun- ment. RT-PCR products were resolved on 2% agarose/TAE gels, wald). For sequence comparison, a noggin gene from the pufferfish blotted onto Hybond-Nϩ membranes (Amersham), and hybridized Fugu rubripres was cloned from a lambda library (gift from Sam to confirm the nature of the amplified fragments. For follistatin and Aparicio). noggin, the entire inserts of the plasmids pBS-zfol3Ј and pCS2- zfnog were labeled and hybridized using the Church system as Northern Blotting described above. The EF1␣ PCR fragment was detected by hybridization with the internal oligonucleotide AGGAGGGTAAT- Embryos were staged according to Westerfield (1994) and stage- GCTAGCG end-labeled with T4 polynucleotide kinase (Boeh- specific total RNA was isolated using Trizol LS reagent (Gibco ringer) as recommended by the supplier. Hybridization was carried BRL) according to the supplier’s instructions. For Northern analy- out overnight at 37°C in 6ϫ SSC, 5ϫ Denhardt’s solution, 1% SDS sis, 10 ␮g of total RNA per lane was dissolved in 1ϫ Mops, 12% with washes at 25°C in 2ϫ SSC, 0.1% SDS. formaldehyde, and 80 ␮g/ml ethidium bromide, then supplemented with an equal volume of formamide, denatured at 65°C for Linkage Analysis 5 min, chilled on ice, supplemented with 1/4 volume of loading dye (30% Ficoll, 0.1% bromophenol blue, and 0.1% xylene cyanole), To assess whether the mercedes mutation is genetically linked loaded on a gel (1% agarose, 0.75% formaldehyde in 1ϫ Mops), and to the noggin gene, linkage analysis was carried out using a

single-strand conformation polymorphism (SSCP) which allowed After several washes in PBS (Sigma) supplemented with 0.1% us to distinguish the noggin gene of the mes mutant strain mestz209 Tween 20, embryos were dechorionated and stored in 100% metha- (Hammerschmidt et al., 1996a) from that of the wik wild-type nol until needed. Whole-mount in situ hybridization with reference line (Rauch et al., 1997). A mestz209/mestz209 homozygous digoxygenin-labeled RNA probes and antibody staining were essen- mutant fish was crossed to a wik fish (P0), the F1 offspring were tially done according to Schulte-Merker et al. (1992), with slight raised and crossed with each other, and the resulting F2 offspring modifications described in Hammerschmidt et al. (1996a). For were sorted according to mes or wild-type phenotypes. Genomic double in situ RNA-immunostainings, embryos underwent a stan- DNA from zebrafish adults and pools of 15 mutant or wild-type dard in situ hybridization, followed by a fixation for 2–6 h in 4% embryos was isolated as described in Westerfield (1994). PCR paraformaldehyde/PBS and a standard immunostaining. For pho- amplification of a 209-bp fragment of the noggin coding region was tography, stained embryos were cleared in benzylbenzoate:benzyl carried out as described for RT-PCR analysis except that 25 cycles alcohol:xylol 10:5:1. The xylol was added because morphological were performed, using the primers GACCGAACTCAGAGCCAT- structures are better visible compared to conventional mounting in TCTAGGC (sense) and GACCAGAGCCACAGCTGAAGCC (an- benzylbenzoate:benzyl alcohol. Photographs were taken on Kodak tisense). Gel electrophoresis and sample preparation were carried 64T slide film under a Zeiss Axiophot microscope, and slides were out as described previously (Kishimoto et al., 1997, and reference scanned with a Nikon LS-1000 slide scanner and processed using cited therein), with the exception that the Pharmacia clean gel kit Photoshop 4.0 software. was used according to the supplier’s instructions. While the P0 fish showed single bands which were clearly distinguishable from each other, wild-type and mutant F2 pools gave both bands, indicating RESULTS that the noggin gene and the mestz209 mutation are not linked. Cloning of follistatin and noggin from Zebrafish Formation of Synthetic RNA and Injection A low-stringency screening approach was chosen to into Embryos clone the zebrafish homologue of the follistatin gene. By The zebrafish follistatin full-length cDNA was amplified and a screening zebrafish gastrula-stage libraries, we obtained Kozak sequence ACC (Kozak, 1986) was introduced upstream of two independent full-length clones of follistatin with a the start codon, using the oligonucleotides CGCGATCCTC- coding region of 966 bp which gives rise to a protein of GAGACCATGCTAAGGATGCTAAAGC and GCGGATCCT- 321 amino acids. As in mammals and Xenopus, two forms TACTTACAGTTGCAAGATCCTGAATGC as sense and anti- of follistatin transcripts appear to be present in zebrafish. sense primers and the full-length follistatin clone pBS-fol as Northern hybridization reveals two distinct RNAs of template DNA. After BamHI digest, this fragment was cloned into approximately 3.5 and 2.4 kb (Fig. 2B). These two tran- the BglII site of pSP64TS (Krieg and Melton, 1984) to give the transcription construct p64TS-fol. For the generation of the ze- scripts might result from alternative splicing, as has been brafish noggin transcription vector pCS2-zfnog, a SalI fragment of demonstrated for follistatin in pig, human, and rat. In plasmid pXEX.znoggin (S.S. and R.M.H., unpublished) was cloned human, the longer transcript is caused by partial splicing into the XhoI site of pCS2 (Rupp et al., 1994). For synthesis of of intron 5 when an internal splice acceptor site is used. capped mRNA, p64TS-fol was linearized with EcoRI, pCS2-zfnog This longer transcript gives rise to a shorter, carboxy- with NotI, and the Xenopus follistatin and chordin transcription truncated form of Follistatin protein which lacks a vectors pXF-64TNE (Hemmati-Brivanlou et al., 1994) and pSP35T- stretch of 27 mostly acidic amino acids encoded by the chd (Sasai et al., 1994) with XbaI followed by in vitro transcription last exon (Shimasaki et al., 1988). In pig, this shorter form with SP6 RNA polymerase using the Ambion Message Machine kit has been reported to be a stronger inhibitor of Activin according to the supplier’s instructions. Sense RNA encoding than the longer version (Sugino et al., 1993). Consistent Xenopus Noggin (Xnog; pSP64␤m-noggin; Smith et al., 1992), Xenopus Bmp4 (Xbmp4; pSP64T-xBmp4; Ko¨ster et al., 1991; Dale with the much higher abundance of the 3.5-kb transcript et al., 1992; Jones et al., 1992), and a truncated Bmp receptor (tBr; in the zebrafish, all three of our independently isolated pSP64T-tBr; Graff et al., 1994; Maeno et al., 1994; Suzuki et al., zebrafish follistatin cDNA clones appear to correspond to 1994) was generated as previously described (Hammerschmidt et the splicing variant which gives rise to the shorter al., 1996a,b). Different amounts of mRNA were injected into the protein, as revealed by sequence comparison with fol- yolk of one- to four-cell-stage embryos, as previously described listatin subtypes of other species (see Fig. 1A for a (Hammerschmidt et al., 1996c) comparison of the short versions of zebrafish and Xeno- pus Follistatin with the long version of mouse Follista- Fish Maintenance, Photography, in Situ tin). The N-terminal region of zebrafish Follistatin (Fig. Hybridization, Immunostaining, and Sectioning 1A) contains a putative signal peptide (von Heijne, 1986) consisting of 2 basic amino acids (KR), a central hydro- Fish were maintained and bred as described in Mullins et al. phobic region (ILLLFWLCYLI), and a more polar region (1994). As wild-type strain, the TL line was used (Haffter et al., (EDQKVQ) anterior to the putative cleaving site (A/G), 1996). noggin in situ hybridizations were carried out with the albino line albb4. For mutant analyses, the chordino allele dintt250 indicating that Follistatin is secreted. The major part of and the swirl/bmp2b allele swrta72 were used. Photographs of live the mature proteins consists of three closely related embryos were taken after anesthetization with tricaine (Wester- domains that have been described as Follistatin modules field, 1994) and mounting in 5% methylcellulose. For staining, (Patthy and Nikolics, 1993). Here, the sequences show embryos were fixed overnight at 4°C in 4% paraformaldehyde/PBS. particularly high conservation. In general, the proteins

with noggin genomic clones from other vertebrates, the noggin coding region contains no introns (McMahon et al., 1998), so genomic fragments were used for synthesis of mRNAs. Noggin has diverged more extensively between fish and tetrapods than has Follistatin (Fig. 1B). Here, the mouse and Xenopus proteins are 76% identical, while zebrafish Noggin has 56 and 55% of the amino acid residues in common with the respective Xenopus and mouse proteins. Despite intensive searches using low-stringency hybridizations of genomic and cDNA libraries and polymerase chain reactions with degenerated primers, however, no other noggin-related genes were found in zebrafish or pufferfish F. rubripres (see above).

Zebrafish Follistatin and Noggin Have Dorsalizing Activities In Xenopus embryos, Noggin and Follistatin have been shown to induce neural specification (Lamb et al., 1993; Hemmati-Brivanlou et al., 1994) and to dorsalize ventral mesoderm (Smith and Harland, 1992; Smith et al., 1993; Fainsod et al., 1997). We have previously demonstrated that Xenopus Noggin displays similar neuralizing and mesoderm-dorsalizing activities when overexpressed in zebrafish embryos (Hammerschmidt et al., 1996b). Here, we carried out the same tests for zebrafish noggin and zebrafish and Xenopus follistatin (Fig. 2). Injection of mRNA of any of the three genes leads to a general dorsalization of wild-type zebrafish embryos in a dose-dependent fashion (Fig. 2; Table 1). noggin appears much more effective than follistatin. In contrast to noggin, even the highest amounts of follistatin RNA used did not lead to the strongest dorsalized phenotype (C5, Table 1). To achieve more moderate phenotypes (C4–C1), approximately 25 times more follistatin RNA than noggin RNA was required (Table 1). No significant difference in dorsalizing efficiencies, however, was detectable between the zebrafish and the Xenopus homologues of noggin and follistatin (Table 1). Depending on the amount FIG. 1. Deduced amino acid sequence and alignment of zebrafish of injected RNA, the entire phenotypic range of the various Follistatin (A) and Noggin (B) with the homologous Xenopus and dorsalized mutants could be mimicked, as shown here for mouse proteins. Identical amino acids are shaded black. For ze- zebrafish follistatin injections (Figs. 2A–2E). Low amounts brafish Follistatin, the putative signal sequence, determined ac- of mRNA caused a phenotype found in minifin mutants cording to von Heijne (1986), and the three Follistatin modules (Fig. 2A), and intermediate amounts gave a piggy tail-like (Patthy and Nikolics, 1993) are marked. The alignment has been phenotype (Fig. 2B), while the maximal amounts caused a performed with the longer splice variant of mouse Follistatin. Noggin proteins (B) are more distantly related to each other than more severe phenotype, like that of snailhouse or weak Follistatins (see text for more detail). Abbreviations: Z, zebrafish; swirl mutants (Figs. 2C and 2D; Mullins et al., 1996). The X, Xenopus; M, mouse; fol, Follistatin; nog, Noggin. phenotypes result from a general expansion of dorsal fates at the expense of ventral fates during late blastula and gastrula stages, as revealed by molecular markers. In strongly affected embryos, the expression of the ventral from Xenopus and mouse are more closely related to each marker gene eve1 (Joly et al., 1993) in the presumptive other (81% identical amino acids) than is zebrafish Fol- mesoderm of the marginal zone is strongly reduced (not listatin to both of them (79% sequence identity with shown), while in the presumptive ectoderm in animal Xenopus and 74% with mouse Follistatin). regions, the expression domain of the neuroectodermal A similar approach was taken to clone zebrafish noggin, marker gene fkd3 (J. Odenthal, in preparation; Hammer- but since noggin sequences were not detected in several schmidt et al., 1996a) is enlarged, extending into ventral- cDNA libraries, it was cloned from a genomic library. As most regions of the injected embryo (Figs. 2I–2L). This

FIG. 2. Functional analysis of zebrafish follistatin and noggin. Both molecules can dorsalize zebrafish embryos. (A–D) Overexpression of zebrafish and Xenopus follistatin in wild-type embryos leads to a dose-dependent dorsalization. Weakly dorsalized embryos (C1) display a partial loss of the ventral tail fin, as indicated in (A) by an arrow. Moderate dorsalization (C3) is characterized by a wound-up tail, the “piggy tail” (Mullins et al., 1996) phenotype (B), while in strongly dorsalized embryos (C4/C5), the entire embryonic axis is wound up in a snailshell-like fashion (C, D). (E–H) chordino mutant embryos can be rescued by overexpression of zebrafish follistatin (F, G). Note the normalization of eye sizes and blood islands (indicated by arrows in F–H). All embryos in (A–H) are shown at approximately 36 hpf. (I–L) Expression of fkd3; 70% epiboly, animal view, dorsal right; in the wild-type embryo. (I) fkd3 expression is confined to the dorsal side, while it is expanded into ventralmost regions in embryos injected with zebrafish follistatin (J), Xenopus follistatin (K), or zebrafish noggin (L). Abbreviations: ϩ, wildtype; hpf, hours after fertilization; din,chordino mutant dintt250. The second abbreviation in the upper right corner indicates the injected mRNA: fol, zebrafish follistatin; Xfol, Xenopus follistatin; nog, zebrafish noggin.

shows that zebrafish Follistatin and Noggin can dorsalize Injection of appropriate amounts of follistatin mRNA ventral mesoderm and induce neural specification in early into dintt250 mutant embryos leads to a significant nor- zebrafish embryos, as has been previously reported for their malization of blood volume and the size of eyes and head Xenopus homologues. (Figs. 2E–2H; see Hammerschmidt et al., 1996b, for Not only are Follistatin and Noggin able to dorsalize noggin). This suggests that zebrafish noggin and follista- wild-type embryos, they also can rescue chordino mu- tin could substitute for at least part of the chordino D-V tants which bear a null mutation in the zebrafish chordin patterning function if they were expressed like their gene (Schulte-Merker et al., 1997; Fisher et al., 1997). Xenopus homologues.

TABLE 1 follistatin and noggin Misexpression in Wild-Type and chordino Mutant Embryos

No. %C5 %C4 %C3 %C2 %C1 % %V1 %V3 Exp n (swr) (snh) (pgy) (laf) (mfn) WT (mes) (din)

Xenopus noggin in WT 50 pg/embryo 1 57 100 0000000 10 pg/embryo 1 39 100 0000000 4 pg/embryo 2 98 46 53100000 Zebrafish noggin in WT 10 pg/embryo 1 87 100 0000000 4 pg/embryo 2 120 29 57 11 3 0 1 0 0 Xenopus follistatin in WT (100 pg/embryo) 3 269 0 55 22 15 7 1 0 0 Zebrafish follistatin in WT (100 pg/embryo) 3 212 0 38 26 10 12 15 0 0 lacZ in WT (100 pg/embryo) 1 137 0 000010000 Xenopus follistatin in dintt250 (100 pg/embryo) 1 69 0 99100000 Zebrafish follistatin in dintt250 (100 pg/embryo) 1 88 0 21 20 16 7 27 9 0 lacZ in dintt250 (100 pg/embryo) 1 63 0 000073027

Note. No. Exp, number of experiments; n, number of scored embryos. C1–C5 represent dorsalized phenotypes of increasing strength, as described in Mullins et al. (1996) and Kishimoto et al. (1997). C5 is the strongest phenotype, comparable to that found in swirl (swr) mutants, characterized by football-shaped embryos at the end of gastrulation and lysis of the embryos around the 15-somite stage. C4 to C1 were defined based on the phenotype at 36 hpf. C4 embryos are characterized by a body axis which is wound up in a snailshell-like fashion, as in snailhouse (snh) mutant embryos (see Figs. 2C and 2D). In C3 embryos, the tail is wound up, while the trunk is normal, as in piggy tail (pgy) mutants (see Fig. 2B). C2 is characterized by a complete loss of the ventral tail fin and a curled-up tip of the tail, like in lost a fin (laf) mutants. C1 embryos display a partial loss of the ventral tail fin, while the shape of the entire embryo is normal, comparable to the phenotype of minifin (mfn) mutants (see Fig. 2A). The ventralized V3 phenotype of chordino (din) mutants is characterized by a small head and small eyes, loss of posterior notochord, enlarged blood islands, and multiple ventral tail fins (Fig. 2H; Hammerschmidt et al. 1996a), while in V1 embryos, all of these traits are much more subtle; the eyes and the head are of almost normal size, the blood islands are slightly enlarged, and the ventral tail fin displays a weak duplication at its posterior end, similar to the characteristics of mercedes (mes) mutant embryos (Figs. 2F and G; Hammerschmidt et al., 1996a). For the chordino rescue experiments, embryos from a cross of two dintt250/ϩ heterozygous fish were used. These crosses normally lead to the chordino mutant phenotype in 25% of embryos, hpf, hours after fertilization, WT, wild-type.

Temporal Expression Pattern of Zebrafish continuing throughout all examined later stages, with a follistatin and noggin maximum at the 8- and 15-somite stages (Figs. 3A and 3B). This expression pattern clearly contrasts with that of the In Xenopus, expression starts for both noggin and follista- D-V patterning genes chordino and bmp2b and bmp4, tin prior to the onset of gastrulation (Smith et al., 1992; whose expression is initiated much earlier, during blastula Hemmati-Brivanlou et al., 1994). In zebrafish, we examined and early gastrula stages. Expression of chordino starts at the temporal expression pattern of noggin by developmen- the dome stage, as revealed by parallel RT-PCR analysis tal RT-PCR and RNase protection analysis and the expres- (not shown), and expression of bmp2b and bmp4 at the sion pattern of follistatin by RT-PCR and Northern analy- sphere and the shield stage, respectively (Nikaido et al., sis. In contrast to their homologues in Xenopus embryos, no 1997). zebrafish noggin or follistatin transcripts could be detected in early gastrula (shield) or younger zebrafish embryos via RT-PCR after 20 (Fig. 3A) or 40 amplification cycles (not noggin and bmp2/4 Are Expressed in a Mutually shown). However, low levels of noggin expression were Exclusive Fashion in Jaw Arches, Neurocranium, found at 80% epiboly and throughout later phases of gas- and Pectoral Fin Buds trulation and segmentation, with a marked increase at day The spatial expression patterns of zebrafish noggin and 1 and maximal levels at day 2 of development (Fig. 3A). follistatin were investigated by whole-mount in situ hy- Similar results were obtained by RNase protection analysis bridization. Although low levels of noggin transcription (not shown). For follistatin, RT-PCR and Northern analysis were detected via RT-PCR, we failed to find localized both revealed expression beginning at 80% epiboly and noggin transcripts via in situ hybridization of gastrula or

arches and the ethmoid plate and the trabeculae cranii of the neurocranium and in the entire inner region of the pectoral fin buds (Figs. 4A and 4B). At this stage, expression of both bmp2b and bmp4 is very weak and diffuse (not shown). At 2.5 days of development, noggin expression has faded in the central zone of the trabeculae cranii and in the basihyal of the hyoid arch (Fig. 2C), while a complementary region of each structure expresses bmp4 (Fig. 2D). At 3 days of development, noggin expression is prominent in the basibranchial and ceratobranchial components of the gill arches (Figs. 2E and 2F), while no bmp2b or bmp4 expression can be detected in these structures at this stage of development. By 3.5 days of development, however, noggin expression in the gill arches has almost completely ceased, while strong bmp2b expression has appeared in the ceratobranchial regions (Fig. 2H). Thus, although difficult to compare, since they are expressed at slightly different stages, the expression domains of bmp2b and noggin in the gill arches appear complementary. noggin shows expression in the central basibranchial region and in four bilateral stripes in the ceratobranchial region, while bmp2b expression is restricted to five bilateral ceratobranchial stripes. In pectoral fin buds of 2-day-old larvae, noggin shows a broad expression in mesenchymal cells (Fig. 4B), while after 3 days of development, noggin expression appears to be confined to a one cell diameter thick plate, in which it is expressed in a graded fashion; expression starts to fade in distal regions of its expression domain, coincident with FIG. 3. Temporal expression pattern of zebrafish follistatin and noggin. (A) Developmental RT-PCR analysis for follistatin, noggin, alterations in the shape of cells which acquire a and the ef1␣ (control). Neither noggin nor follistatin transcripts are chondrocyte-typical, honey-comb-like appearance (Fig. 4J), detected before 80% epiboly. noggin signals are weak during late while strong noggin expression is maintained in proximal gastrulation and segmentation stages, but much stronger during regions of the fin buds (Figs. 4I and 4J). At all investigated primordia stages, with a maximum at day 2. follistatin signals are stages (day 2 through day 3.5), the noggin expression do- of comparable strength throughout all investigated stages from main is surrounded by bmp2b- and bmp4-positive mesen- 80% epiboly onward, with a slight maximum at the 8- and chymal and/or epithelial cells in the distalmost regions of 15-somite stages. (B) Developmental Northern blot hybridized with the fin bud (Figs. 4H and 4I). follistatin probe. Ten micrograms of stage-specific total RNA was loaded per lane and stained with ethidium bromide as a loading control. Consistent with the results of RT-PCR, expression is first follistatin Expression in Head Mesoderm, detectable at 80% epiboly, reaches a maximum at day 1 of Somites, Eyes, and Brain development, and then declines. Note the two transcripts of follistatin. The larger transcript is 3.5 kb, the shorter transcript The first distinct follistatin transcripts are detectable at approximately 2.4 kb. 60% epiboly in dispersed presumptive mesodermal cells located in paraxial hypoblast regions animal to the marginal zone (Figs. 5A and 5B). In addition, in some, but not all in situ hybridizations, we observed a few follistatin-positive segmentation stage embryos except for two spots of expres- cells in the dorsal midline (indicated by an arrow in Figs. 5A sion near the tailbud of late-stage gastrulae (not shown). At and 5D; see also Fig. 7B). These cells were always located in primordia stages, noggin expression was observed in the jaw posterior axial regions of gastrulation embryos, while no region of the head and in the pectoral fin buds (Fig. 4). In staining at all was seen in the presumptive prechordal plate mouse embryos, the equivalent structures have been re- region in anterior axial positions (Fig. 5E; see also Figs. 6D, ported to express Bmp2 and Bmp4 (Hogan, 1996), the 7A, and 7B), a region characterized by the expression of putative noggin antagonists during Xenopus D-V pattern organizer-specific genes like goosecoid (Figs. 5C and 5D; formation. Therefore, we carried out parallel in situ hybrid- Stachel et al., 1993; Schulte-Merker et al., 1994b; Thisse et izations of sibling embryos for zebrafish noggin,bmp2b, and al., 1994) and chordino (Fig. 7F; Schulte-Merker et al., 1997; bmp4 (Nikaido et al., 1997; Martıńez-Barbera´ et al., 1997) Miller-Bertoglio et al., 1997). As gastrulation proceeds, the transcripts. At 2 days of development, noggin is strongly paraxial follistatin expression domain appears to grow (Figs. expressed in presumptive cartilage cells of the pharyngeal 5C and 5D); at the end of gastrulation, it covers about a

FIG. 4. Expression pattern of noggin and bmp2b and bmp4 in branchial arches and pectoral fin primordia, revealed by whole mount in situ hybridization of albino zebrafish larvae. (A and B) 48 hpf, noggin expression, lateral view (A) and dorsal view (B) on head. noggin displays a broad expression in inner regions of the pectoral fin buds (pfb), while cells at the surface lack noggin expression. In addition, noggin is

Copyright © 1998 by Academic Press. All rights of reproduction in any form reserved. Follistatin and Noggin in Zebrafish 497 quarter of the dorsal side of the embryo with a sharp neuroectodermal derivatives. Compared to the mesoderm, posterior boundary and a more diffuse anterior boundary at however, expression in the neuroectoderm is initiated the level of the anterior end of the notochord anlage, rather late, between the 5- and the 10-somite stage (Figs. 5G marked by expression of no tail (Schulte-Merker et al., and 5H). At the 20-somite stage, follistatin is expressed in 1992, 1994a) (Fig. 5E). To determine the exact position of the eye vesicles, with strongest expression in posterior the posterior border of follistatin expression, we carried out regions of the forming cups (Fig. 5K). In the brain, it is double in situ hybridization together with myoD (Weinberg expressed in four distinct stripes of the forebrain, and in et al., 1996), a marker for adaxial cells of the presomitic mid- and hindbrain, while the region of the midbrain– mesoderm (Fig. 5F). The follistatin and myoD expression hindbrain boundary, which is characterized by the expres- domains partly overlap, indicating that the paraxial follista- sion of engrailed (Hatta et al., 1991), lacks follistatin tin expression comprises not only the region which gives expression (Fig. 5K). In addition to the anteroposterior rise to the head mesoderm, but also the anterior part of the pattern, there seems to be a specific dorsoventral pattern of presomitic mesoderm. Cells of this presomitic region give follistatin expression, as is for instance evident in cross- rise to the first 4 somites and continue to express follistatin sections through anterior regions of the hindbrain (Fig. 5M). during somite formation and later segmentation stages In addition, the banded expression in the forebrain is (Figs. 5G and 5H), while in more posterior somites, follista- restricted to ventral regions, while no expression at all is tin expression is initiated approximately 2 h after somites detected in dorsal regions at the same anteroposterior are formed. In somite 6, for instance, no follistatin tran- positions of the brain (not shown). In the eye vesicles, broad scripts are detectable at the 10-somite stage (Fig. 2H), but retinal expression is found ventrally, while in more dorsal they are seen at the 12-somite stage (Fig. 2I). Expression in regions, follistatin expression is restricted to the outer layer the posterior somites is initiated in a specific spatiotempo- of the forming cups, the stratum pigmenti, from which the ral pattern. It starts in ventral regions of the somites that pigment epithelium will be formed (Fig. 5L). most likely undergo sclerotomal specification (Fig. 5P; compare twi expression in Hammerschmidt et al., 1996c). Early follistatin Expression Is Regulated by In addition, follistatin starts to be expressed in dorsal Bmp2b, Chordin, and Follistatin somitic cells in close proximity to the neural tube, from whence the expression progressively spreads over the entire We have focused on early follistatin expression in the myotome (Fig. 5P) with the exception of adaxial regions anterior paraxial mesoderm, carrying out injection and adjacent to the notochord, regions characterized by the mutant analyses to determine the extent to which this early expression of eng (Hatta et al., 1991) which remain expression is regulated by molecules previously shown to follistatin-negative in all somites (Figs. 5I and 5O). There play important roles during dorsoventral patterning, may also be a connection between the neural tube and namely Bmp2/4, Chordin, and Follistatin itself. For the follistatin expression in the anterior paraxial mesoderm. In mutant analyses, we examined the follistatin expression the first 4 somites and the posterior head mesoderm, where pattern in zebrafish embryos that lack either a functional the paraxial region is rather small and close to the neural chordino gene (dintt250; Schulte-Merker et al., 1997) or a tube, follistatin expression is maintained in all mediolat- functional bmp2b gene (swrta72; Kishimoto et al., 1997). In eral positions (except the adaxial positions of the somites, wild-type embryos at the 80–90% epiboly stage, the see above, Fig. 5O). However, in more anterior regions, paraxial follistatin expression domain spans approximately follistatin expression is restricted to an approximately 150° of the circumference of the embryo (Fig. 6A). This three-cell-diameter broad band surrounding the neural tube. dorsolateral extent of the follistatin expression domain is In addition to in mesoderm, follistatin is expressed in significantly altered in chordino and bmp2b mutant em-

strongly expressed in the ethmoid plate (et) and the trabeculae cranii (tc), components of the neurocranium, and in specific subregions of the pharyngeal arches, including the anteriormost portion of the hyoid, the basihyale (bh). Weak staining is also observed in the otic vesicles (ov). (C and D) 60 hpf, noggin expression (C) and bmp4 expression (D), ventrolateral view on head; noggin and bmp4 are expressed in complementary patterns in subregions of the trabeculae cranii and the hyoid. noggin transcripts are present in distal regions of the tc and posterior regions of the hyoid, bmp4 transcripts in central regions of the tc (indicated with short arrows) and the bh (indicated with long arrows). (E and F) 72 hpf, noggin expression, lateral view (E) and dorsal view (F); noggin is expressed in the basibranchial (bb) and ceratobranchial (cb) components of the gill arches, while no expression is detected in the pharyngeal arches. (G and H) 84 hpf, noggin (G) and bmp2b (H) expression, dorsal view; noggin expression in the gill arches has declined, while bmp2b displays strong expression in the basibranchial components of the gill arches and in apical regions of the pectoral fin primordia. (I and J) noggin (I, J) and bmp4 (I) expression in pectoral fin primordia (pfb); (I) 60 hpf, (J) 72 hpf, anterolateral view; noggin (n) is expressed in proximal regions, bmp4 (b) in distal regions of the pectoral fin primordia (I). Expression of noggin in inner cells declines in a distal-to-proximal wave. Cells with reduced noggin levels display an elongated cell shape, which is characteristic for chondrocytes (J). Abbreviations: bb, basibranchial; bh, basihyale; cb, ceratobranchial; et, ethmoid plate; ov, otic vesicle; pfb, pectoral fin bud; tc, trabeculae cranii; hpf, hours after fertilization.

FIG. 5. Expression pattern of follistatin in wild-type zebrafish embryos, revealed by whole-mount in situ hybridization. All embryos except the right embryo in (I) are stained for follistatin transcripts. In the case of double in situ hybridization immunostainings (A, B, E, I–K, M–P) and double in situ hybridizations (C, D, F), the second detected gene product is indicated in the lower right corner. (A and B) 60% epiboly, double stained for No tail protein (brown), dorsal view (A) and close-up of same embryo in dorsolateral view (B); scattered follistatin expression is seen in anterior paraxial regions and occasionally in single cells (indicated by an arrow, also in D) in posterior axial positions within the no tail expression domain in the notochord anlage (A). Expression of follistatin is restricted to involuted cells of the hypoblast (B). (C, D, and E) 80% epiboly (C), 90% epiboly (D), and tailbud stage (E), double stained for goosecoid (gsc) RNA (C, D, indicated with g) and No tail protein (E, brown), dorsal view; the paraxial follistatin expression domain grows, but expression remains excluded from anterior axial regions (E) which are characterized by the expression of the organizer-specific gene goosecoid (C, D). Posterior axial follistatin staining is weaker than staining in paraxial regions. In addition, the intensity of the axial follistatin staining varies in specimens from different in situ hybridizations (compare, e.g., C, D, and Fig. 7B). The reason for this variability is currently unknown. (F) 2-somite stage, double stained for myoD RNA (m), dorsal view; the posterior region of the follistatin expression domain and the anterior region of the myoD expression domain overlap, indicating that follistatin is expressed in the presumptive head mesoderm and a distinct anterior region of the presomitic mesoderm. The follistatin expression in the anterior presomitic mesoderm appears slightly condensed. (G and H) 5-somite stage (G) and 10-somite stage (H), dorsal view; the posterior paraxial follistatin expression is organized in a segmented pattern defining four stripes (indicated by asterisks) at identical anteroposterior positions in embryos of both stages. No additional stripes are visible in the 10-somite embryo. Note the intense anterior staining in (H), representing expression in the brain anlage. (I) 12-somite stage, stained for follistatin

FIG. 6. Expression of follistatin in chordino and bmp2b mutants and after overexpression of Xenopus follistatin,bmp4, or chordin. All embryos shown are at 90% epiboly and are viewed from the animal pole, except in (D and H), which are lateral views. Expression of follistatin is reduced in the chordino mutant embryo (B), spanning the dorsalmost 30° of the embryonic circumference, compared to 150° in wild-type sibling embryos (A). A similar dorsal retraction of the follistatin expression domain is obtained in wild-type embryos after injection of synthetic Xenopus bmp4 mRNA (C). In the bmp2b mutant embryo (E), the follistatin expression domain is expanded, spanning the entire embryonic circumference. The same effect is achieved in wild-type embryos by the injection of Xenopus chordin (F) and Xenopus follistatin mRNA (G). In (H), the Xfol-injected embryo of (G) is shown in a lateral view in comparison to an uninjected wild-type sibling embryo (D), showing that ectopic ventral zebraﬁsh follistatin expression is restricted to the same anteroposterior levels like the wild-type expression on the dorsal side. Abbreviations: ϩ, wildtype; din, chordino mutant dintt250; swr, bmp2b mutant swrta72. The second abbreviation in the upper right corner indicates the injected mRNA: Xbmp4, Xenopus bmp4; Xchd, Xenopus chordin; Xfol, Xenopus follistatin.

bryos. In embryos deﬁcient for the dorsalizer Chordino, 6B). The same effect is achieved by overexpression of the follistatin expression is restricted to dorsalmost paraxial Chordino antagonist Bmp4 (Fig. 6C). In contrast, embryos regions, spanning 30° of the embryonic circumference (Fig. mutated in the bmp2b gene (Kishimoto et al., 1997) display

RNA and No tail protein (left) and for myoD RNA and No tail protein (right), dorsal view on spread embryos. In the left embryo, the anterior tip of the notochord anlage is indicated by an arrow. Comparison of the follistatin and myoD expression patterns assigns the intense segmented follistatin expression to the first 4 somites (indicated by asterisks). In addition, weaker follistatin expression is visible in somites 5 to 7. In all somites, follistatin expression is excluded from adaxial regions. (J) 20-somite stage, double stained for No tail protein, lateral view. Intense follistatin staining is observed in the head and anterior trunk region, weaker staining in more posterior somites. Indicated are the positions for which sections of similar embryos are shown in (L–P). (K) 20-somite stage, double stained for Engrailed protein; neuroectodermal follistatin expression is confined to the eye vesicles and four distinct bands in the forebrain, midbrain, and hindbrain, while the midbrain–hindbrain boundary region, which is characterized by the expression of engrailed, lacks follistatin transcripts. (L–O) Saggital (L) and transverse (M–P) 7-␮m paraffin sections of doubly stained embryos of the 16-somite (M, L) and 25-somite (O, N, P) stage at the respective positions indicated in (J). (L) Section through the ventral (upper eye) and the medial (lower eye) part of the eye vesicles, revealing follistatin expression in posterior and distal regions of the eye primordia. (M) At the level of the anterior hindbrain, paraxial follistatin expression is restricted to an approximately three cell diameter broad region surrounding the neural tube. In the brain, follistatin is expressed in a distinct dorsal band and a more diffuse band in the ventral part of the tube. (N) More posteriorly, in the notochord region anterior of the somites, follistatin is expressed in the narrow band of cells located between the neural tube and the yolk and in ventral paraxial cells. (O) In the first 4, relatively small somites, follistatin is expressed throughout the entire somitic region except the adaxial positions adjacent to the notochord. Note the weak diffuse follistatin expression in the neural tube. (P) In posterior somites, follistatin expression is confined to ventral regions, which most likely represent the presumptive sclerotome, and to dorsal regions of the myotome in the vicinity of the neural tube. Abbreviations: eng, Engrailed; g, gsc,goosecoid;m,myoD, ntl, No tail.

FIG. 7. Comparison of zebrafish follistatin, chordino, bmp2b, and bmp4 expression (A–G) revealed by in situ hybridization and interaction between follistatin and bmp4 (H). All embryos are of the 80% epiboly stage. The in situ probes used are indicated in the lower right corner. (A and B) follistatin expression, lateral view (A), dorsal view (B). (E and F) chordino expression, (E) lateral view, (F) dorsal view; while chordino is expressed in organizer-derived cells in the presumptive prechordal plate, follistatin is expressed in more posterior regions of the notochord anlage. The paraxial chordino and follistatin expression domains in epiblast and hypoblast, respectively, are almost congruent, as confirmed by anti-Ntl counterstaining, marking the notochord anlage and the germ ring as reference points (not shown). At slightly later stages, the paraxial chordino expression domain appears slightly larger than the follistatin domain, extending more into marginal regions (not shown). (C) bmp2b expression, lateral view; the dorsal extension in the ventral expression domain is indicated by an asterisk, the expression in dorsal marginal cells by an arrow. (G) follistatin and bmp2b expression, lateral view. (D) bmp4 expression, lateral view; the expression domain in the dorsal marginal cells is indicated by an arrow, the expression domain in the presumptive prechordal plate by an arrowhead. (H) bmp4 expression after overexpression of zebrafish follistatin, lateral view. The region with reduced bmp4 expression is delimited by two arrows.

a ventral extension of the follistatin expression domain, a dorsalizing signal and a Bmp antagonist during later which spans the entire circumference of the mutant em- phases of D-V patterning, we compared the expression bryos (Fig. 6E). The same extension is achieved after inac- patterns of follistatin and chordino with those of bmp2b tivation of Bmps via a dominant-negative, truncated BMP and bmp4 in midgastrula embryos. At 80% epiboly, follista- receptor (not shown; Graff et al., 1994; Maeno et al., 1994; tin and chordino are expressed in different domains on the Suzuki et al., 1994; Hammerschmidt et al., 1996b) or by the dorsal side of the embryo; chordino is expressed in the overexpression of any of the Bmp2/4 antagonists chordin anterior region of the axis, which derives from the orga- (Fig. 6F), noggin (not shown), or follistatin itself (Fig. 6G). In nizer, while the axial expression of follistatin is restricted any of these treatments, ectopic ventral follistatin expres- to more posterior regions (Figs. 7B and 7F). In paraxial sion remains restricted to the anteroposterior levels of the regions, chordino and follistatin are expressed at similar wild-type follistatin expression domain, and no ectopic anteroposterior levels (Figs. 7B and 7F); however, follistatin follistatin expression is detected in the marginal zone or in expression is restricted to the mesoderm, while chordino is the animalmost regions (Figs. 6D and 6H). largely expressed in the neuroectoderm (Miller-Bertoglio et al., 1997; S. Schulte-Merker and M.H., in preparation). bmp2b and bmp4 transcripts are found on both the follistatin and bmp2/4 Are Expressed in a Mutually ventral and the dorsal side of the gastrulating embryo. Exclusive Fashion Dorsally, they are transiently expressed at the dorsal mar- As described above, Follistatin has dorsalizing activities gin in involuting presumptive mesodermal cells (Figs. 7C and is expressed—although not in the organizer itself—in and 7D). In addition, bmp4 is expressed in the anterior dorsal and paraxial regions of the gastrulating zebraﬁsh dorsal mesoderm of the prechordal plate, a derivative of the embryo. In order to investigate whether it might function as organizer (Fig. 7D). The axial follistatin expression domain

Copyright © 1998 by Academic Press. All rights of reproduction in any form reserved. Follistatin and Noggin in Zebrafish 501 is complementary to those of bmp2b and bmp4 and is DISCUSSION located in the notochord anlage between the two dorsal bmp2b/bmp4 expression sites (compare Fig. 7A with Figs. Zebrafish Noggin and Follistatin Do Not 7C and 7D). Contribute to Organizer Function On the ventral side, bmp2b is more broadly expressed than bmp4 (Figs. 7C and 7D). The dorsolateral extent of the The goal of the present study was to investigate the roles bmp2b expression domain varies along the anteroposterior of Follistatin and Noggin during early dorsoventral pattern axis; it displays a dorsal extension halfway between the formation of the zebrafish embryo. In Xenopus, Follistatin animal pole and the marginal zone (Fig. 7C). This extension and Noggin are generated by the Spemann organizer. To- is faced by the ventral border of the paraxial follistatin gether with other secreted organizer proteins, Chordin, expression domain (Fig. 7G), suggesting that Follistatin Xnr3, and Cerberus, they may dorsalize surrounding tissues might act to restrict Bmp activity in this specific region of by blocking the ventralizing Bone Morphogenetic Proteins the embryo. This possibility is supported by the effect of Bmp2 and Bmp4. In the zebrafish, mutants with potential overexpressed follistatin on the bmp2b and bmp4 expres- null mutations in the chordin gene, named chordino (origi- sion pattern. Injection of follistatin RNA leads to a selec- nally dino), were isolated. The strong ventralization ob- tive loss of bmp4 transcripts in medial anteroposterior served in mutant embryos indicates that chordino is an regions, while the expression in marginal and animal re- essential component of the fish organizer whose function is gions is less or not at all affected (Fig. 7H). not completely redundant to that of other organizer signals. A similar effect is seen in embryos in which Chordin protein is degraded by the overexpression of the Chordin- follistatin and noggin Are Not Linked to specific protease Xolloid/Tolloid (Blader et al., 1997; Pic- Ventralizing Zebrafish Mutations colo et al., 1997). However, chordino mutants retain some As described above, follistatin, and maybe also noggin, neuroectoderm and some dorsolateral mesoderm in regions could be involved in later phases of dorsoventral patterning close to the organizer, indicating that there must be an- of the zebrafish embryo. To test whether they might be other, most likely a short-range, signaling molecule with essential for these processes, we investigated whether any inductive properties similar to those of Chordino. Of mutations obtained in the large-scale mutagenesis screens course, Noggin and Follistatin would be good candidates for (Driever et al., 1996; Haffter et al., 1996) that affect D-V this additional signal. polarity involve follistatin or noggin. Assuming a dorsaliz- We have isolated zebrafish homologues of follistatin and ing function of the genes, loss-of-function mutations in noggin. The deduced amino acid sequences of both ze- follistatin or noggin should lead to a ventralized phenotype. brafish genes are similar enough to those from Xenopus and Ventralized mutants isolated during the screens define only mouse to be considered their structural homologues, espe- two genes required for dorsal development, namely dino cially since we failed to find other related genes. Like their (Hammerschmidt et al., 1996a), which as mentioned is Xenopus counterparts, the zebrafish genes show strong identical to the zebrafish chordin gene (Schulte-Merker et dorsalizing activities upon misexpression in zebrafish em- al., 1997), and mercedes (Hammerschmidt et al., 1996a), bryos and can rescue the chordino mutant phenotype. which appears to be allelic to the ogon mutation isolated in However, unlike the Xenopus genes, neither noggin nor the Boston screen (Solnica-Krezel et al., 1996; A.M. and follistatin is expressed at the right time or the right place to M.H., unpublished) and whose molecular nature is not yet contribute to organizer function. Expression of neither gene known. To investigate whether the mercedes phenotype is detectable at early gastrula stages, when chordino mu- might be caused by mutations in the zebrafish noggin gene, tants display first signs of the remaining neural induction linkage analysis between mes and noggin was performed, (Hammerschmidt et al., 1996a). This indicates that the taking advantage of a SSCP which allowed us to distinguish zebrafish organizer must contain dorsalizing and neuraliz- a noggin PCR fragment of the mestz209 strain from that of a ing signals other than the described Noggin, Chordino, and wik wild-type line used for the linkage cross (see Materials Follistatin proteins. Such an additional signal might be and Methods). Analysis of DNA from F2 hybrid embryos encoded by mercedes, a second gene required for zebrafish revealed the presence of the wik-specific noggin allele in organizer function which has been identified in the large- mes mutant embryos, indicating that the mutation causing scale mutant screen whose molecular nature is not yet the mes phenotype and the noggin gene are not linked (not known. Linkage analysis between the noggin gene and the shown). Furthermore, no missense or nonsense mutation mercedes mutation and genomic mapping of mercedes and was found in the coding region of the noggin gene that had follistatin showed that neither of the two genes is identical been amplified and sequenced from genomic DNA of ho- to mercedes. mozygous mutant adult mestm305 and mestz209 fish. In Are there other candidates for mercedes and the addi- addition, zebrafish follistatin and mercedes were mapped tional dorsalizing and neuralizing signal present in the relative to simple sequence length polymorphism genetic zebrafish organizer of chordino-deficient zebrafish em- markers (Knapik et al., 1996) and found to reside in two bryos? It is possible that the zebrafish genome contains different linkage groups (Pascal Haffter, pers. comm.). additional, maybe more distantly related, follistatin-or

Copyright © 1998 by Academic Press. All rights of reproduction in any form reserved. 502 Bauer et al. noggin-like genes which in contrast to those described here riorly than chordino and follistatin in involuting marginal are expressed in the zebrafish organizer region. In addition, cells that give rise to the notochord. This expression stops other candidates have been identified in Xenopus, whose shortly after cells have finished involution, simultaneously zebrafish homologues have not been described as yet. The with the initiation of follistatin expression, suggesting that Nodal-related protein Xnr3, a member of the TGF␤ family, Follistatin may be involved in the inhibition of Bmps in the is present in the Xenopus organizer region and can act both notochord anlage. In paraxial positions, chordino and fol- as a dorsalizer of ventral mesoderm (Smith et al., 1995) and listatin are expressed in the region along the presumptive as a direct neural inducer (Hansen et al., 1997). Several anteroposterior axis where bmp2b expression extends far- findings suggest that Xnr3—like Chordin—functions via thest into the dorsal side of the midgastrula embryo. Thus, the inhibition of Bmps. However, this inhibition appears to it is possible that especially here, Bmp antagonists are occur at the Bmp synthesis or secretion step or at some required to suppress Bmp action after its initial attenuation downstream point like receptor binding (Hansen et al., by organizer signals. Indeed, the expression domain of 1997). Taken together, these data would make a zebrafish bmp4, which depends on bmp2b (Kishimoto et al., 1997) homologue of Xnr3 a strong candidate for a partner of and thus could be regarded as a readout of Bmp2b action, Chordino to antagonize Bmp signaling in and around the lacks the dorsal extension of the bmp2b expression domain. zebrafish organizer. Cerberus might be another candidate, This suggests that Bmp2b activity in this region may have since it binds and blocks Bmp activity (Bouwmeester et al., been blocked by paraxial Chordino and Follistatin. 1996; Hsu et al., 1998). However, Cerberus is relatively A further ventral retraction of bmp4 expression at these much less active that Noggin, and it is likely to be primarily specific anteroposterior levels was observed after follistatin an antagonist of the mesoderm-inducing Nodal relatives overexpression. This differential sensitivity of bmp4 ex- (Hsu et al., 1998). Other Bmp antagonists related to Cerbe- pression to Follistatin could indicate a higher competence rus have been discovered (Hsu et al., 1998), and although of cells to Bmp blockage, possibly because these cells are they are not expressed in the Xenopus organizer, it is more dependent on Bmp autoregulation (Kishimoto et al., possible that their expression, like that of noggin and 1997) than cells in marginal regions and close to the animal follistatin, has diverged between zebrafish and Xenopus, so pole. No reduction at all was detected for bmp4 expression that they have an organizer function in the fish. Recently, in the presumptive prechordal plate, even after maximal zebrafish Fgf8, a member of the fibroblast growth factor overexpression of the more potent dorsalizer noggin, indi- family, which is present in a dorsolateral gradient along the cating that bmp4 expression is differently regulated at this marginal zone of late blastula and early gastrula zebrafish site. embryos, was reported to function upstream of Bmp2 and The regulation of follistatin expression itself has been Bmp4, and thereby independently of Chordino, to promote investigated by overexpression studies and mutant analy- dorsal development (Fu¨rthauer et al., 1997). Linkage analyses. These studies show that the dorsoventral extent of the sis of these and other genes, a search for novel signaling follistatin expression domain is regulated by the antagoniz- molecules, and/or positional cloning of mercedes might ing activities of the ventralizer Bmp2b on the one side and reveal the molecular nature of the additional essential Chordino and Follistatin itself on the other, while the organizer signals. initial induction of follistatin expression and the anteroposterior extent of its expression domain appear to depend on other factors. The altered follistatin expression patterns in Follistatin Might Function during Later Phases of mutant embryos also provide some additional insights into D-V Patterning the function of Bmp2b and Chordino during dorsoventral We have shown that follistatin expression starts at mid- patterning of the zebrafish embryo, revealing that the two gastrula stages in presumptive mesodermal cells in anterior signals pattern the anterior mesoderm, thereby determining paraxial and more posterior axial positions. This expression the extent of head mesoderm specification. pattern is largely complementary to that of chordino. In the axis, chordino is expressed more anteriorly than follistatin Follistatin, a General Inhibitor of TGF␤ Proteins? in cells of the presumptive prechordal plate, while in paraxial regions, follistatin and chordino expression is In addition to its expression during gastrulation, follista- confined to different germ layers; follistatin is expressed in tin is expressed at multiple sites during later stages of the mesoderm, chordino in the neuroectoderm and in development, including head mesoderm, somites, eyes, and comparably few mesodermal cells close to the axis. This brain. At the same stages, bmp2b and bmp4 are expressed differential expression indicates that Follistatin and in various tissues like dorsal ectoderm, lateral plate, fin Chordino function in different dorsal domains of mid- and primordia, sclerotome of somites, head mesenchyme, eyes, late gastrula zebrafish embryos, which may be important, and olfactory and otic placodes (Chen et al., 1997; Nikaido since their antagonists Bmp2 and Bmp4 are involved in et al., 1997; H.B. and M.H., unpublished observations). At patterning both ectoderm and mesoderm (Sasai et al., 1995; some of these sites, the bmps are expressed in close prox- Hammerschmidt et al., 1996b). imity to follistatin. In the eye vesicles, for instance, bmp4 In the axis, bmp2b and bmp4 are expressed more poste- is expressed in dorsal regions of the retina (Nikaido et al.,

1997), while follistatin expression is confined to ventral ized by the expression of the homeobox gene engrailed retinal regions and the pigmental zone. A similar, mutually (Ekker et al., 1992; Hatta et al., 1991). A similar strictly exclusive expression pattern is found in the region of the mutually exclusive expression of follistatin and engrailed otic placodes (H.B. and M.H., unpublished observation). has been found in the midbrain–hindbrain border region. Thus, Follistatin could also act to inhibit Bmp proteins Engrailed proteins have been reported to function as tran- during later stages of development. In addition, it could scriptional repressors (Han and Manley, 1993), and it will be function as an inhibitor of other members of the TGF␤ interesting to investigate whether the patterned expression superfamily like Activins, as whose antagonist it was of follistatin in somites and brain of zebrafish embryos is originally described (Nakamura et al., 1989). activin ␤B achieved by the repressing action of Engrailed transcription (Wittbrodt and Rosa, 1994), for instance, is expressed in the factors. midbrain–hindbrain boundary region, perfectly complementary to the expression of follistatin (H.B. and M.H., noggin and Chondrogenesis unpublished observation). In addition, radar, another TGF␤ protein (Rissi et al., 1995), is expressed in dorsal regions of For noggin, we found localized expression at two major the retina, similar to bmp4 and complementary to follista- sites, the head skeleton and the pectoral fins. In the fin buds tin, suggesting that Follistatin may here inhibit both Bmp4 of 2-day-old larvae, noggin is expressed in a large domain and Radar activity. Furthermore, follistatin is expressed at encompassing almost the entire interior of the bud, irre- several sites of zebrafish embryos without any apparent spective of the ongoing different specifications of mesen- correlation to bmp2b, bmp4, activin ␤B or radar expression chymal cells in this region. A subset of the noggin-positive (e.g., in the forebrain), suggesting that Follistatin may here cells is for instance characterized by simultaneous expres- block other not yet investigated TGF␤ proteins. Likewise, sion of both sna1 (Hammerschmidt and Nu¨sslein-Volhard, comparative analyses of the spatiotemporal expression pat- 1993) and myoD (Weinberg et al., 1996), which identifies tern of follistatin and Activins in mouse embryos has led to them as muscle precursors, while other cells of the noggin the hypothesis of additional functions for Follistatin in expression domain will give rise to cartilage. At day 2.5 of addition to inhibition of Activins (Albano et al., 1994; development, continued noggin expression is detected in a Feijen et al., 1994). one cell diameter thick plate of cells which are identified as chondrocytes by their typical honey-comb-like cell shape. noggin is also prominently expressed in the mouse at all Differential follistatin Expression in Somites sites of cartilage condensation from mesenchyme, and its During gastrulation and early somitogenesis stages of expression persists until cartilage becomes hypertrophic zebrafish development, follistatin is continuously ex- and is replaced by bone. Noggin function is required for pressed in the region of the first five somites, but not farther proper regulation of cartilage growth and joint formation caudally, where expression first comes up 2 to 3 h after (Brunet et al., 1998). This aspect of expression in the somite formation. A similar specific expression of follista- developing cartilage may reflect the ancestral state, and tin in the precursors of the anteriormost somites, although noggin may have been co-opted later for expression in the with a different time course, has also been described in the organizer and axial mesoderm of tetrapods. mouse (Albano et al., 1994). This may be interesting, since During the course of pectoral fin bud development in the striking differences in the development and the behavior of zebrafish, noggin expression declines in a distal-to-proximal anterior and posterior somites have recently been reported wave, with strongest expression maintained in proximal- in various vertebrate embryos. In Xenopus, the rostral five most cells at the base of the bud (see Figs. 7B, 7F, 7I, and 7J). somites were found to be formed due to the action of a late This contrasts the proximal-to-distal direction of cartilage blastula organizer, whereas more posterior somites are built formation in the chicken limb bud, where continued pro- up under the influence of a gastrula or trunk organizer liferation of undifferentiated mesenchyme occurs at the tip located animally of the late blastula organizer (Lane and of the limb (Saunders, 1948). Future experiments are needed Keller, 1997). In zebrafish, several mutants in which the to address whether the noggin expression pattern reflects formation of posterior somites is impaired, while anterior some crucial differences between the architecture of fish fin somites develop normally, have been isolated (van Eeden et buds and amniote limb buds. As in mouse (Hogan, 1996), al., 1996). In addition, anterior and posterior somites were however, bmp2b and bmp4 are expressed in distalmost reported to differ in their response to pertussis toxin, an mesenchymal cells in the border region of the developing inhibitor of inhibitory G proteins which may be involved in zebrafish fin buds, complementary to the expression of the regulation of Sonic hedgehog signaling (Hammer- noggin. schmidt and McMahon, 1998). Future studies will have to Comparative expression analyses also indicate a mutu- show how the differential somitic follistatin expression and ally exclusive expression of noggin and bmp2/4 in the jaw the specific features of the first five somites correlate. and the branchial arches. In several examined arches, Another striking feature of somitic follistatin expression bmp2b or bmp4 expression comes up shortly after noggin we observed in zebrafish embryos is its exclusion from expression has ceased, suggesting that Noggin may function adaxial regions of the somites. These regions are character- as a negative regulator of bmp2/4 expression. In the mouse,

Copyright © 1998 by Academic Press. All rights of reproduction in any form reserved. 504 Bauer et al. increased Bmp activity increases the recruitment of cells and where which of the proteins is used during the various into cartilage, as does the loss of Noggin activity in noggin- processes of vertebrate development. deficient mice (Brunet et al., 1998; McMahon et al., 1998). However, there is no evidence in mouse for upregulation of bmp expression in cartilage that lacks noggin (Brunet et al., ACKNOWLEDGMENTS 1998), whereas at other sites, such as the posterior noto- We are very grateful to Drs. D. Melton, J. Graff, and O. Kelly for chord or ES cells in culture, bmp expression is upregulated the transcription plasmids pSP64T-xBmp4, pSP64T-tBr, and pXF- in the absence of noggin (Johansson and Wiles, 1995; 64TNE encoding Xenopus Follistatin, Xenopus Bmp4, and a trun- McMahon et al., 1998). cated dominant version of the Xenopus Bmp2/4 receptor, respec- Due to temporal exclusion of bmp2/4 and noggin expres- tively; to Drs. Y. Sasai and E. M. De Robertis for the Xenopus sion in the zebrafish, it is not clear whether the genes are chordin transcription vector pSP35T-chd; to Dr. S. Schulte-Merker expressed in the same or in neighboring cells. From the for the chordino in situ probe and anti-Ntl antibody; to Dr. Eric comparison of the expression pattern of noggin with that of Weinberg for the myoDin situ probe; to Dr. C. Dreyer for anti- other markers like goosecoid or twist (Hammerschmidt, Engrailed 4D9 antibody; and to Drs. D. Grunwald and J. Campos- Ortega for zebrafish cDNA libraries. Special thanks are due to Dr. 1994), it appears that noggin is expressed in presumptive Pascal Haffter for the information that follistatin and mercedes cartilage cells themselves. This would be consistent with map to different linkage groups and to H. Grandl for his comments its expression in Xenopus (Smith et al., 1992) and mouse on the noggin expression patterns. (Brunet et al., 1998). In amniotes, some bmps are expressed in and close to developing cartilage (Duprez et al., 1996; reviewed in Hogan, 1996). Thus, bmp and noggin expres- REFERENCES sion can, in principle, overlap. Albano, R. M., Arkell, R., Beddington, R. S. P., and Smith, J. C. (1994). Expression of inhibin subunits and follistatin during Evolutionary Aspects of Follistatin, Noggin, and postimplantation mouse development: Decidual expression of Chordin Function activin and expression of follistatin in primitive streak, somites and hindbrain. Development 120, 803–813. As described above, organizer-specific expression of chor- Beddington, R. S. P. (1994). Induction by a second neural axis by the din, follistatin, and noggin is different in different verte- mouse node. Development 120, 613–620. brate organisms. chordin is expressed in the organizer Blader, P., Rastegar, S., Fischer, N., and Stra¨hle, U. (1997). Cleavage region of all investigated vertebrate embryos (Xenopus, of the BMP-4 antagonist Chordin by zebrafish Tolloid. Science Sasai et al., 1994; chicken, Streit et al., 1998; zebrafish, 278, 1937–1940. Schulte-Merker et al., 1997; Miller-Bertoglio et al., 1997). Bouwmeester, T., Kim, S. H., Sasai, Y., Lu, B., and DeRobertis, E. M. (1996). Cerberus is a head-inducing secreted factor ex- noggin transcripts, however, are found in the organizer in pressed in the anterior endoderm of Spemann’s organizer. Nature mouse, chick, and Xenopus (Smith et al., 1992; Connolly et 382, 595–601. al., 1997; McMahon et al., 1998), but not in zebrafish, while Brunet, L. J., McMahon, J. A., McMahon, A. P., and Harland, R. M. follistatin transcripts are present in the organizer in Xeno- (1998). Noggin, cartilage morphogenesis, and joint formation in pus (Hemmati-Brivanlou et al., 1994), but absent from the the mammalian skeleton. Science 280, 1455–1457. organizer in mouse (Albano et al., 1994; Feijen et al., 1994) Chen, J.-N., Van Eeden, F. J. M., Warren, K. S., Chin, A., Nu¨sslein- and zebrafish. Furthermore, they display some crucial dif- Vollhard, C., Haffter, P., and Fishman, M. C. (1997). Left–right ferences in their expression pattern outside the organizer pattern of cardiac BMP4 may drive asymmetry of the heart in region. While in Xenopus, chordin expression is restricted zebrafish. Development 124, 4373–4382. to the organizer and its derivatives, chordino in zebrafish is Church, G. M., and Gilbert, W. (1984). Genomic sequencing. Proc. Natl. Acad. Sci. USA 81, 1991–1995. also expressed in the neuroectoderm, starting at midgas- Connolly, D. J., Patel, K., and Cooke, J. (1997). Chick noggin is trula stages (Miller-Bertoglio et al., 1997; S. Schulte-Merker expressed in the organizer and neural plate during axial develop- and M.H., in preparation). During early somite stages, ment, but offers no evidence of involvement in primary axis chordino shows a segmented expression in various stripes formation. Int. J. Dev. Biol. 41, 389–96. in the hindbrain region (Miller-Bertoglio et al., 1997; S. Dale, L., Howes, G., Price, B. M. J., and Smith, J. C. (1992). Bone Schulte-Merker and M.H., in preparation), like follistatin in morphogenetic protein 4: A ventralizing factor in early Xenopus the mouse (Albano et al., 1994; Feijen et al., 1994), but development. Development 115, 573–585. unlike follistatin in the zebrafish. De Robertis, E. M., and Sasai, Y. (1996). A common plan for In contrast to the variable expression patterns of chordin, dorsoventral patterning in Bilateria. Nature 380, 37–40. noggin, and follistatin in different vertebrate species, the Dosch, R., Gawantka, V., Delius, H., Blumenstock, C., and Niehrs, C. (1997). Bmp4 acts as a morphogen in dorsoventral patterning expression patterns of their putative antagonists, the bmp in Xenopus. Development 124, 2325–2334. genes, appear more conserved. Thus, assuming that the Driever, W., Solnica-Krezel, L., Schier, A. F., Neuhauss, S. C. F., common function of Chordin, Noggin, and Follistatin as Malicki, J., Stemple, D. L., Stanier, D. Y. R., Zwartkruis, F., inhibitors of Bmps, and possibly of other TGF␤ molecules, Abdeliah, S., Rangini, Z., Belak, J., and Boggs, C. (1996). A genetic has been conserved throughout evolution, it appears that screen for mutations affecting embryogenesis in zebrafish. De- different “plans of action” have evolved, regulating when velopment 123, 37–46.

Duprez, D., Bell, E. J. D. H., Richardson, M. K., Archer, C. W., Hansen, C. S., Maison, C. D., Steele, K., George, S., and Smith, Wolpert, L., Brickell, P. M., and Francis-West, P. H. (1996). W. C. (1997). Direct neural induction and selective inhibition of Overexpression of BMP-2 and BMP-4 alters the size and shape of mesoderm and epidermis inducers by Xnr3. Development 124, developing skeletal elements in the chick limb. Mech. Dev. 57, 483–492. 145–157. Harland, R. M., and Gerhart, J. C. (1997). Formation and function of Ekker, M., Wegner, J., Akimenko, M. A., and Wersterfield, M. Spemann’s organizer. Annu. Rev. Cell. Dev. Biol. 13, 611–667. (1992). Coordinate embryonic expression of three zebrafish en- Hatta, K., Bremiller, R., Westerfield, M., and Kimmel, C. B. (1991). grailed genes. Development 116, 1001–1010. Diversity of expression of engrailed-like antigens in zebrafish. Fainsod, A., Deißler, K., Yelin, R., Marom, K., Epstein, M., Pille- Development 112, 821–832. mer, G., Steinbeisser, H., and Blum, M. (1997). The dorsalizing Hemmati-Brivanlou, A., Kelly, O. G., and Melton, D. A. (1994). and neural inducing gene follistatin is an antagonist of BMP-4. Follistatin, an antagonist of activin, is expressed in the Spemann Mech. Dev. 63, 39–50. organizer and displays direct neuralizing activity. Cell 77, 283– Feijen, A., Goumans, M. J., and van den Eijnden-van Raaij, A. J. M. 295. (1994). Expression of activin subunits, activin receptors and Hogan, B. L. M. (1996). Bone morphogenetic proteins: Multifunc- follistatin in postimplantation mouse embryos suggests specific tional regulators of vertebrate development. Genes Dev. 10, development functions for different activins. Development 120, 1580–1594. 3621–3637. Holley, S., Neul, J. L., Attisano, L., Wrana, J. L., Sasai, Y., Fisher, S., Amacher, S. L., and Halpern, M. E. (1997). Loss of O’Connor, M. B., De Robertis, E. M., and Ferguson, E. L. (1996). cerebrum function ventralizes the zebrafish embryo. Develop- The Xenopus dorsalizing factor Noggin ventralizes Drosophila ment 124, 1301–1311. embryos by preventing DPP from activating its receptor. Cell 86, Fu¨rthauer, M., Thisse, C., and Thisse, B. (1997). A role for FGF-8 in 607–617. the dorsoventral patterning of the zebrafish gastrula. Develop- Holley, S. A., and Ferguson, E. L. (1997). Fish are like flies are like ment 124, 4253–4264 frogs: Conservation of dorsal–ventral patterning mechanisms. Graff, J. M., Thies, R. S., Song, J. J., Celeste, A. J., and Melton, D. A. BioEssays 19, 281–284. Holley, S. A., Jackson, P. D., Sasai, Y., Lu, B., De Robertis, E. M., (1994). Studies with a Xenopus BMP receptor suggest that ventral Hoffmann, F. M., and Ferguson, E. L. (1995). A conserved system mesoderm-inducing signals override dorsal signals in vivo. Cell for dorsal–ventral patterning in insects and vertebrates involving 79, 169–179. sog and chordin. Nature 376, 249–253. Haffter, P., Granato, M., Brand, M., Mullins, M. C., Hammer- Hsu, D. R., Economides, A. N., Wang, X., Eimon, P. M., and schmidt, M., Kane, D. A., Odenthal, J., van Eeden, F. J. M., Jiang, Harland, R. M. (1998). The Xenopus dorsalizing factor Gremlin Y.-J., Heisenberg, C.-P., Kelsh, R. N., Furutani-Seiki, M., Warga, identifies a family of secreted proteins that antagonize BMP R. M., Vogelsang, E., Beuchle, D., Schach, U., Fabian, C., and activities. Mol. Cell. 1, 673–683. Nu¨sslein-Volhard, C. (1996). The identification of genes with Johansson, B. M., and Wiles, M. V. (1995). Evidence for involve- unique and essential functions in the development of the ze- ment of activin A and bone morphogenetic protein 4 in mam- brafish, Danio rerio. Development 123, 1–36. malian mesoderm and hematopoietic development. Mol. Cell. Hammerschmidt, M. (1994). “Isolierung und Charakterisierung Biol. 15, 141–151. von Genen und Mutanten zur Analyse Fru¨hembryonaler Muster- Joly, J.-S., Joly, C., Schulte-Merker, S., Boulkebache, H., and Con- bildung im Zebrafish, Brachydanio rerio.” Eberhard-Karls- damine, H. (1993). The ventral and posterior expression of the Universita¨t, Tu¨ bingen. [Ph.D. thesis] homeobox gene eve1 is perturbed in dorsalized and mutant Hammerschmidt, M., and Nu¨sslein-Volhard, C. (1993). The expres- embryos. Development 119, 1261–1275. sion of a zebrafish gene homologous to Drosophila snail suggests Jones, C. M., Lyons, K. M., Lapan, P. M., Wright, W. V. E., and a conserved function in invertebrate and vertebrate gastrulation. Hogan, B. L. M. (1992). DVR-4 (bone morphogenetic protein-4) as Development 119, 1107–1118. a posterior ventralizing factor in Xenopus mesoderm induction. Hammerschmidt, M., Pelegri, F., Mullins, M. C., Kane, D. A., van Development 115, 639–647. Eeden, F. J. M., Granato, M., Brand, M., Furutani-Seiki, M., Jones, C. M., and Smith, J. C. (1995). Revolving vertebrates. Curr. Haffter, P., Heisenberg, C.-P., Jiang, Y.-J., Kelsh, R. N., Odenthal, Biol. 5, 574–576. J., Warga, R. M., and Nu¨sslein-Volhard, C. (1996a). dino and Kishimoto, Y., Lee, K.-H., Zon, L., Hammerschmidt, M., and mercedes, two genes regulating dorsal development in the ze- Schulte-Merker, S. (1997). The molecular nature of swirl: BMP2 brafish embryo. Development 123, 95–102. function is essential during early dorsoventral patterning. Devel- Hammerschmidt, M., Serbedzija, G. N., and McMahon, A. P. opment 124, 4457–4466. (1996b). Genetic analysis of dorsoventral pattern formation in Knapik, E. W., Goodman, A., Atkinson, O., Roberts, C. T., the zebrafish: Requirement of a BMP-like ventralizing activity Shiozawa, M., Sim, C. U., Weksler-Zangen, S., Trolliet, M. R., and its dorsal repressor. Genes Dev. 10, 2452–2461. Futrell, C., Innes, B. A., Koike, G., McLaughlin, M. G., Pierre, L., Hammerschmidt, M., Bitgood, M. J., and McMahon, A. P. (1996c). Simon, J. S., Vilallonga, E., Roy, M., Chiang, P.-W., Fishman, Protein kinase A is a common negative regulator of Hedgehog M. C., Driever, W., and Jacob, H. J. (1996). A reference cross DNA signaling in the vertebrate embryo. Genes Dev. 10, 647–658. panel for zebrafish (Danio rerio) anchored with simple sequence Hammerschmidt, M., and McMahon, A. P. (1998). The effect of length polymorphisms. Development 123, 451–460. pertussis toxin on zebrafish development: A possible role for Ko¨ster, M., Plessow, S., Clement, J. H., Lorenz, A., Tiedemann, H., inhibitory G-proteins in Hedgehog signaling. Dev. Biol. 194, and Kno¨chel, W. (1991). Bone morphogenetic protein 4 (BMP-4), 166–171. a member of the TGF-␤ family, in early embryos of Xenopus Han, K., and Manley, J. L. (1993). Functional domains of the laevis: Analysis of mesoderm inducing activity. Mech. Dev. 33, Drosophila engrailed protein. EMBO J. 12, 2723–2733. 191–200.

Kozak, M. (1986). Point mutations define a sequence flanking the Inhibition of ventral signals by direct binding of Chordin to AUG initiator codon that modulates translation by eukaryotic BMP-4. Cell 85, 589–598. ribosomes. Cell 44, 283–292. Piccolo, S., Agius, E., Lu, B., Goodman, S., Dale, L., and DeRobertis, Krieg, P. A., and Melton, D. A. (1984). Functional messenger RNAs E. M. (1997). Cleavage of Chordin by Xolloid metalloprotease are produced by SP6 in vitro transcription of cloned cDNAs. suggests a role for proteolytic processing in the regulation of Nucleic Acids Res. 12, 7057–7070. Spemann organizer activity. Cell 91, 407–416. Lamb, T. L., Knecht, A. K., Smith, W. C., Stachel, S. E., Econo- Rauch, G.-J., Granato, M., and Haffter, P. (1997). A polymorphic mides, A. N., Stahl, N., Yancopolous, G. D., and Harland, R. M. zebrafish line for genetic mapping using SSLPs on high- (1993). Neural induction by the secreted polypeptide noggin. percentage agarose gels. Technical Tips Online, T01208. Science 262, 713–718. Rissi, M., Wittbrodt, J., De`lot, E., Naegeli, M., and Rosa, F. M. Lane, M.C., and Keller, R. (1997). Microtubule disruption reveals (1995). Zebrafish radar: A new member of the TGF-␤ superfamily that Spemann’s organizer is subdivided into two domains by the defines dorsal regions in neural plate and the embryonic retina. vegetal alignment zone. Development 124, 895–906 Mech. Dev. 49, 223–234. Lemaire, P., and Kodjabachian, L. (1996). The vertebrate organizer: Rupp, R. A. W., Snider, L., and Weintraub, H. (1994). Xenopus Structure and molecules. Trends Genet. 12, 525–531. embryos regulate the nuclear localization of XMyoD. Genes Dev. Maeno, M., Ong, R. C., Suzuki, A., Ueno, N., and Kung, H. (1994). 8, 1311–1323. A truncated bone morphogenetic protein receptor alters the fate Sasai, Y., Lu, B., Steinbeisser, H., Geissert, D., Gont, L. K., and De of ventral mesoderm to dorsal mesoderm: Roles of animal pole Robertis, E. M. (1994). Xenopus chordin: A novel dorsalizing tissue in the development of the ventral mesoderm. Proc. Natl. factor activated by organizer-specific homeobox genes. Cell Acad. Sci. USA 91, 10260–10264. 79, Martıńez-Barbera´, J. P., Toresson, H., Da Rocha, S., and Krauss, S. 779–790. (1997). Cloning and expression of three members of the zebrafish Sasai, Y., Lu, B., Steinbeisser, H., and De Robertis, E. M. (1995). Bmp family: Bmp2a,Bmp2b and Bmp4. Gene 198, 53–59. Regulation of neural induction by the Chd and Bmp-4 antagonis- Matzuk, M. M., Lu, N., Vogel, H., Sellheyer, K., Roop, D. R., and tic patterning signals in Xenopus. Nature 376, 333–336. Bradley, A. (1995). Multiple defects and perinatal death in mice Saunders, J. W. (1948). The proximo-distal sequence of origin of the deficient in follistatin. Nature 374, 360–363. parts of the chick wing and the role of the ectoderm. J. Exp. Zool. McMahon, J. A., Takada, S., Zimmerman, L. B., Fan, C.-M., 108, 363–404. Harland, R. M., and McMahon, A. P. (1998). Noggin mediated Schulte-Merker, S., Ho, R. K., Herrmann, B. G., and Nu¨sslein- antagonism of BMP signaling is required for growth and pattern- Volhard, C. (1992). The protein product of the zebrafish homo- ing of the neural tube and somite. Genes Dev., 12, 1438–1452. logue of the mouse T gene is expressed in nuclei of the germ ring Melton, D. A. (1984). Efficient in vitro synthesis of biologically and the notochord of the early embryo. Development 116, active RNA and RNA hybridization probes from plasmids con- 1021–1032. taining a bacteriophage SP6 promoter. Nucleic Acids Res. 12, Schulte-Merker, S., Eeden, F. V., Halpern, M. E., Kimmel, C. B., and 7035. Nu¨sslein-Volhard, C. (1994a). No tail (ntl) in the zebrafish Miller-Bertoglio, V. E., Fisher, S., Sańchez, A., Mullins, M. C., and homologue of the mouse T (Brachyury) gene. Development 120. Halpern, M. E. (1997). Differential regulation of chordin expres- Schulte-Merker, S., Hammerschmidt, M., Beuchle, D., Cho, K. W., sion domains in mutant zebrafish. Dev. Biol. 192, 537–550. DeRobertis, E. M., and Nu¨sslein-Volhard, C. (1994b). Expression Mullins, M. C., Hammerschmidt, M., Haffter, P., and Nu¨sslein- of the zebrafish goosecoid and no tail gene products in wild-type Volhard, C. (1994). Large-scale mutagenesis in the zebrafish: In and mutant ntl embryo. Development 120, 843–852. search of genes controlling development in a vertebrate. Curr. Schulte-Merker, S., Lee, L. J., McMahon, A. P., and Hammer- Biol. 4, 189–202. schmidt, M. (1997). The zebrafish organizer requires chordino. Mullins, M. C., Hammerschmidt, M., Kane, D. A., Odenthal, J., Nature 387, 862–863. Brand, M., van Eeden, F. J. M., Furutani-Seiki, M., Granato, M., Shimasaki, S., Koga, M., Esch, F., Cooksey, K., Mercado, M., Koba, Haffter, P., Heisenberg, C.-P., Jiang, Y.-J., Kelsh, R. N., and A., Ueno, N., Ying, S.-Y., Ling, N., and Guillemin, R. (1988). Nu¨sslein-Volhard, C. (1996). Genes establishing dorsoventral Primary structure of the human follistatin precursor and its pattern formation in the zebrafish embryo: The ventral specify- genomic organization. Proc. Natl. Acad. Sci. USA 85, 4218– ing genes. Development 123, 81–93. 4222. Nakamura, T., Takio, K., Eto, Y., Shibai, H., Titani, K., and Sugino, Smith, W. C., and Harland, R. M. (1992). Expression cloning of H. (1989). Activin-binding protein from rat ovary is Follistatin. noggin, a new dorsalizing factor localized in the Spemann orga- Science 247, 836–838. Nikaido, M., Tada, M., Saij, T., and Ueno, N. (1997). Conservation nizer in Xenopus embryos. Cell 70, 829–840. of BMP signaling in zebrafish mesoderm patterning. Mech. Dev. Smith, W. C., Knecht, A. K., Wu, M., and Harland, R. M. (1993). 61, 75–88. Secreted noggin protein mimics the Spemann organizer in dor- Nordness, S., Krauss, S., and Johansen, T. (1994). cDNA sequence of salizing Xenopus mesoderm. Nature 361, 547–549. zebrafish (Brachydanio rerio) translation elongation factor-1 al- Smith, W. C., McKendry, R., Ribisi, S., and Harland, R. M. (1995). pha: Molecular phylogeny of eukaryotes based on elongation A nodal-related gene defines a physical and functional domain factor-1 alpha protein sequences. Biochim. Biophys. Acta 1219, within the Spemann organizer. Cell 82, 37–46. 529–532. Solnica-Krezel, L., Stemple, D. L., Mountcastle-Shah, E., Rangini, Patthy, L., and Nikolics, K. (1993). Functions of agrin and agrin Z., Neuhauss, S. C. F., Malicki, J., Schier, A., Stanier, D. Y. R., related proteins. Trends Neurosci. 16, 76–81. Zwartkruis, F., Abdelilah, S., and Driever, W. (1996). Mutations Piccolo, S., Sasai, Y., Lu, B., and De Robertis, E. M. (1996). A affecting cell fates and cellular rearrangements during gastrula- possible molecular mechanism for Spemann organizer function: tion in the zebrafish. Development 123, 67–80.

Spemann, H., and Mangold, H. (1924). U¨ ber die Induktion von Jiang, Y.-J., Kane, D. A., Kelsh, R. N., Mullins, M. C., Odenthal, Embryonalanlagen durch Implantation artfremder Organisa- J., Warga, R. M., Allende, M. L., Weinberg, E. S., and Nu¨sslein- toren. Wilhelm Roux’s Arch. Entwicklungsmech. 100, 599–638. Volhard, C. (1996). Mutations affecting somite formation and Stachel, S. E., Grunwald, D. J., and Myers, P. Z. (1993). Lithium patterning in the zebrafish, Danio rerio. Development 123, perturbation and goosecoid expression identify a dorsal specifi- 153–164. cation pathway in the pregastrula zebrafish. Development 117, von Heijne, G. (1986). A new method for predicting signal sequence 1261–1274. cleavage sites. Nucleic Acids Res. 14, 4883–4890. Streit, A., Lee, K. J., Woo, I., Roberts, C., Jessell, T. M., and Stern, C. D. Weinberg, E. S., Allende, L. M., Kelly, C. S., Abdelhamid, A., (1998). Chordin regulates primitive streak development and the Murakami, T., Andermann, P., Doerre, O. G., Grunwald, D. J., stability of induced neural cells, but is not sufficient for neural and Riggleman, B. (1996). Developmental regulation of zebrafish induction in the chick embryo. Development 125, 507–519. myoD in wild-type, no tail and spadetail embryos. Development Sugino, K., Kurosawa, N., Nakamura, T., Takio, K., Shimasaki, S., 122, 271–280. Ling, N., Titani, K., and Sugino, H. (1993). Molecular heteroge- Westerfield, M. (1994). “The Zebrafish Book: A Guide for the neity of Follistatin, an activin-binding protein. J. Biol. Chem. Laboratory Use of Zebrafish.”Univ. of Oregon Press, Eugene. 268, 15579–15587. Wittbrodt, J., and Rosa, F. M. (1994). Disruption of mesoderm and Suzuki, A., Thies, R. S., Yamayi, N., Song, J. J., Wozney, J. M., axis formation in fish by ectopic expression of activin variants: Murakami, K., and Ueno, N. (1994). A truncated bone morphoge- The role of maternal activin. Genes Dev. 8, 1448–1462. netic protein receptor affects dorsal–ventral patterning in the early Zimmerman, L. B., and Harland, R. M. (1996). Bmp-4 function is Xenopus embryo. Proc. Natl. Acad. Sci. USA 91, 10255–10259. blocked by high affinity binding to the Spemann organizer signal Thisse, C., Thisse, B., Halpern, M. E., and Postlethwait, J. H. (1994). Noggin. Cell 85, 599–606. goosecoid expression in neuroectoderm and mesendoderm is disrupted in zebrafish cyclops gastrulas. Dev. Biol. 164, 420–429. Received for publication April 17, 1998 van Eeden, F. J. M., Granato, M., Schach, U., Brand, M., Furutani- Revised June 29, 1998 Seiki, M., Haffter, P., Hammerschmidt, M., Heisenberg, C.-P., Accepted June 29, 1998